Battle-tested prompts for Claude, GPT, and Gemini
Claude 4.5 changed everything. The new model is more literal and precise — old prompts that relied on Claude "filling in the blanks" now underperform. We've added 5 new techniques optimized for 2026:
| Technique | What It Does |
|---|---|
| XML Structured | Structure prompts with tags Claude natively understands |
| Extended Thinking | Force step-by-step reasoning for complex problems |
| Self-Correcting | Built-in verification to catch hallucinations |
| Anti-Botspeak | Avoid phrases that scream "AI wrote this" |
| Permission to Fail | Get honest "I don't know" instead of confident lies |
- Quick Start
- Quick Reference
- Browse by Problem
- Browse by Category
- Prompt Format
- Pro Prompts
- Contributing
# Clone
git clone https://github.com/unisone/ai-prompts.git
# Or copy any prompt directly from GitHubEach prompt includes: Description | When to Use | The Prompt | Why It Works | Example
All 54 prompts at a glance:
| Problem | Solution | Link |
|---|---|---|
| AI makes errors I catch too late | Have it self-critique before finalizing | Self-Critique |
| Complex tasks come back incomplete | Break into sub-problems first | Decomposition |
| AI jumps to wrong solutions | Force context gathering before solving | Context-First |
| Responses are too long/verbose | Explicit output constraints | Concise Output |
| Claude 4.x gives minimal answers | Request explicit scope expansion | Literal Instructions |
| Problem | Solution | Link |
|---|---|---|
| Missing bugs in review | Systematic 10-point checklist | Code Review |
| Debugging takes forever | 8-step root cause methodology | Debugging |
| Afraid to refactor | 5-phase safe refactoring | Refactoring |
| Inconsistent API design | REST/GraphQL design template | API Design |
| Low test coverage | Generate happy path + edge cases | Test Generator |
| Lost in new codebase | 9-part codebase analysis | Explain Codebase |
| Problem | Solution | Link |
|---|---|---|
| Documentation is lacking | README, API docs, runbook templates | Technical Docs |
| Resume not converting | ATS-optimized resume builder | Resume Architect |
| Problem | Solution | Link |
|---|---|---|
| Data but no insights | Patterns, anomalies, actions framework | Data Insights |
8 prompts for development workflows.
| Prompt | Description |
|---|---|
| code-review-checklist.md | 10-point review covering logic, security, performance |
| debug-systematic.md | 8-step methodology to find root cause |
| refactor-safely.md | 5-phase approach: assess, safety net, change, verify, cleanup |
| api-endpoint-design.md | REST/GraphQL endpoint design template |
| test-generator.md | Generate happy path, boundary, error tests |
| explain-codebase.md | Understand unfamiliar codebases systematically |
| ultrathink-code-craftsman.md | Philosophy-driven system prompt |
| structured-reasoning-planner.md | 9-step structured reasoning framework |
5 meta-prompting methods backed by research.
| Prompt | Research Basis |
|---|---|
| self-critique.md | Self-criticism improves accuracy (Schulhoff, Learn Prompting) |
| decomposition.md | Breaking problems into parts improves complex task completion |
| context-first.md | Context is underrated and massively impactful (OpenAI) |
| concise-output.md | Models overcompensate when instructions are vague |
| literal-instructions.md | Claude 4.x takes instructions literally (Anthropic) |
2 prompts for documentation and professional writing.
| Prompt | Description |
|---|---|
| technical-documentation.md | README, API docs, runbook templates |
| ultrathink-resume-architect.md | ATS-optimized resume optimization |
1 prompt for data analysis.
| Prompt | Description |
|---|---|
| data-insights.md | Extract patterns, anomalies, and actionable recommendations |
Every prompt follows this structure:
# Prompt Name
## Description
What this prompt does.
## When to Use
Specific scenarios where this prompt helps.
## Prompt
The actual prompt text to copy/paste.
## Why It Works
Research or reasoning behind the technique.
## Example
Real usage showing input and output.
## Notes
Tips and limitations.These prompts built a 577-file production app with 0 critical bugs.
| Free (This Repo) | Pro (Sponsors Only) |
|---|---|
| 17 production prompts | System prompts I use daily |
| Research-backed techniques | Multi-file refactoring patterns |
| Ship Feature in 2 Hours workflow | |
| Senior code review (catches security issues) | |
| New prompts added monthly |
- Fork the repository
- Create a prompt using the template
- Submit a PR
See CONTRIBUTING.md for details.
Built by Alex Zaytsev
Star this repo if you find it useful
AI Prompts Pro — Premium prompts for sponsors:
| What You Get | Free | Pro |
|---|---|---|
| Basic prompts | ✅ | ✅ |
| System prompts (Claude Code, Cursor) | ❌ | ✅ |
| Enterprise patterns | ❌ | ✅ |
| Production templates (SaaS, CLI) | ❌ | ✅ |
| Parallel agent orchestration | ❌ | ✅ |
| Monthly updates | ❌ | ✅ |